You can use a growing list with your messages if they have a character indicating the beginning and another indicating the end of a message. You can calculate how many bytes you can store, depending on the buffer size of Teensy's serial. Make sure you send batches of messages that fit this size. Send a sort of "done" message from the Teensy to Pd, so that Pd can send the next batch. This is useful in case you're sending too many messages. Perhaps you can make a mechanism that can drop messages in case the list grows exponentially.
You can use [list-extend] from the list-abs library, which is a library of vanilla abstractions for list manipulation.

One thing that comes to mind is that sometimes, certain hardware combinations cannot print or write the byte 13 from the serial line to Pd. In such cases, Serial.println() sends the number 10 twice to Pd instead of sending 13 and 10. Serial.write(13) sends 10 to Pd. Maybe there's something similar happening the other way round. You could try and echo what you send to the Teensy back to Pd, maybe manually, so you don't clog the serial line. If you see something suspicious, it could be causing the error.

your patch is pretty complex to sink in it and really grasp it. I checked the Arduino code, and I could suggest to either use a series of else if statements instead of lots of single if, or a switch-case structure, with included breaks, so that your code doesn't have to check every single case whenever something arrives in the serial buffer. There could be something in some other part of the code though that I can't spot because it is hardware specific. Try increasing the baud rate to 115200, and probably slowing down the rate of messages from Pd, as you say.

Can you share your code and patch? Is the communication duplex? I don't know if that could lead to your issue, but I had a Teensy freezing in the past with duplex communication without setting the order of data transmission. The project was headless, so I didn't see the output error, so I don't know if it was the same.

I'd like to share with you a performance for live coding on the 3dPdModular system. This system is programmed in Pd, and the Code module enables to write Python code live with the Pyo module for DSP, through the [pyo~] external. This performance googles the latitude and longitude coordinates of randomly selected capital cities, and uses these values as MIDI notes to control the frequencies of two oscillators.
By associating cities to sound, it questions our notion of localism.

[iemguts/closebang]

How about [expr~ if((($v1 > $f2) && ($v1 < $f3)), 1, 0)]

@plusch I don't think it's a good idea to manually put stuff in Pd/extra. What's your OS? Don't you have a Documents/Pd/externals folder? Usually deken creates this folder for you, or something along these lines. It's better to put things in your user environment rather than in your system environment.
I don't know about the chaos library. Is it a collection of objects (many .c and .pd_linux, .d_fat, or .dll files, according to your operating system) or is it one monolithic library with one .c file and one executable? In the case of the latter you'll have to import the library, e.g. [declare -path /path/to/chaos -lib chaos].

@jaffasplaffa I second that.

It seems that the cyclone developers assume that you put the cyclone folder in Pd/externals without putting the cyclone folder in Pd's search path, and then load the objects with the cyclone prefix as it is with the objects of this library in their help patches. I actually agree with this approach as it is easy to figure out which library an external belongs to.
Try removing the cyclone folder from Pd's search path but make sure the cyclone folder is in a folder where Pd will look at, externals is a good place for this. The try to load an object like this [cyclone/avg~] and open its help file.

Have you tried the Pyo module for Python? It's a fully-featured DSP module with OSC support. http://ajaxsoundstudio.com/software/pyo/

Don't know how [http_get] works, but you can have a look at PuREST JSON library for HTTP requests.

$0 doesn't work in messages. You have to do this:

[$0 ] <- this is an object
|
[read -resize /path/to/file.aiff $1-array2( <- this is a message

the only solution I can think of is using abstractions and prepending the $0 value to your [send]/[receive] names. Example, [s $0-wireless-conn]/[r $0-wireless-conn]
There will be no conflict between different instances of abstractions as each gets a unique value associated with $0.

It's -open /path/to/patch.pd

@cfry WiringPi is deprecated due to too much pressure from its audience to its developer (actually bullying). I don't think it will be very easy to use it. You can try Python (check Sparkfun's tutorials) or even the [gpio] object (I think that's what it's called) which is by Ivica Ico Bukvic, I think.

What's the framerate of Processing? 30? Why need audio rate if your fps is 30Hz? Just use control rate OSC at that speed.

Not sure if it's relevant, but Roman Heafeli has made a non-blocking [until] abstraction. It's called [nbuntil]. Perhaps it could help with your last example.

What do you mean exactly? You want to open a Pd patch in Python, or you want to run Pd without the GUI and control it from Python?

deken is pretty straight forward. Go to Help->Find Externals and search. Most likely only externals for your system will appear, so yes is the answer to your last question. If it's for the two libraries/objects you've already mentioned, why bother with deken and not just download the binaries for timbreID and compile the other?

I'm no admin, but I don't see a reason why not.